)]}'
{"nova/tests/functional/regressions/test_bug_1806064.py":[{"author":{"_account_id":4690,"name":"melanie witt","display_name":"melwitt","email":"melwittt@gmail.com","username":"melwitt"},"change_message_id":"7d18c4394b318252577817b0afc57e250bf9cfc8","unresolved":false,"context_lines":[{"line_number":33,"context_line":"    scheduling failure, the instance record along with the associated BDMs"},{"line_number":34,"context_line":"    are created in the cell0 database and that is where the \"local delete\""},{"line_number":35,"context_line":"    code in the API finds them to detach the related volumes. In the case of"},{"line_number":36,"context_line":"    the quota fail race, the instance has already been created in a selected"},{"line_number":37,"context_line":"    cell but the BDMs records have not been and are thus not \"seen\" during"},{"line_number":38,"context_line":"    the API local delete and the volumes are left attached to a deleted server."},{"line_number":39,"context_line":""},{"line_number":40,"context_line":"    An additional issue, covered in the test here, is that tags provided when"}],"source_content_type":"text/x-python","patch_set":1,"id":"3f79a3b5_2632b4c0","line":37,"range":{"start_line":36,"start_character":25,"end_line":37,"end_character":43},"updated":"2018-12-05 22:11:54.000000000","message":"Hrm, this reminds me of how we\u0027ve talked about collapsing the two conductor loops into a single loop and just let all the instance-related resources be created \"at the same time\" and upon a recheck failure, we clean up everything including BDMs and tags. And doing that is the right fix for this.\n\n(later) Thinking about it more, that refactor would not be backport friendly, so I think we\u0027re best off with this sort of fix for backporting and have the refactor be master-only.","commit_id":"5d514b33e28964b38aeb42a8dd5b93f3fc8ae239"}]}